In order to better understand your data and the way SVM works is to begin with a linear SVM. This tybe of SVM is interpretable, which means that each of your 41 features has a weight (or 'importance') associated with it after training. You can then use plot3() with your data on 3 of the 'best' features from the linear svm.
Support Vector Machine A more convenient formulation The previous problem is equivalent to min w,b 1 2 ∥w∥2 2 subject to y i(w·x +b) ≥ 1 for all 1 ≤ i ≤ n. Remarks: This is an optimization problem with linear, inequality constraints. This is an implementation of the SVM algorithm. To do this, I solve the dual L1-regularized and kernelized optimization problem via classic QP using CVX and (in the future) via the SMO algorithm.